Short-term Forecasting Using Physics-Guided Machine Learning

Society is strongly affected by extreme weather events and it is expected that this will become more noticeable in the future.

In the case of such weather events rapid updates of the latest forecast from the numerical weather prediction (NWP) model using the latest observational data could be useful especially for early warnings. 

In this project we investigate whether a system taking advantage of machine-learning techniques, that use constraints based on the laws of physics, could be used for this purpose thus considerably speeding up the generation of nowcast products.

 

 

Results for a machine learning system using constraints from physics for Burgers' equation.

Contact: Camiel Severijns